It snowed in Vienna


However, since people who live in Vienna are used to snow, we had a great turnout at the VSTS event.  This was a very savvy crowd and they asked a lot of great questions around both the depth of the source code control system (details about how to do promotions from branches selectively, how to manage multiple workspaces on a machine, etc) as well as some very good questions on our domain specific languages (DSL) strategy.


I took a couple things away from this open session.  First, we should do a better job of describing “how” you’d use the configuration management system, not just what it is.  I’ve asked Buck Hodges and Howie Hilliker (one of our great doc writers) to work on getting some content up about how workspaces work, and how you would use them to accomplish different tasks.  I also pointed a lot of folks who had detailed questions (ok I admit I couldn’t answer all of them) to go post them on the TFS forum. 


Second, there is still a fair bit of confusion around our stance wrt UML2.  I tried to clear it up by making a few points summarized below:

  • We believe in model driven development deeply integrated across the entire software lifecycle from requirements to design to checking in and out code (my example here being if I want to edit a class which implements a web service, I should automajically get the wsdl file, aspx etc checked out as well and when I edit, they should all be in sync)
  • We believe that UML has a lot of good modeling tools.  As with everyone I’ve ever talked with on the subject, we think there are some that are more valuable than others.
  • We believe that UML is trying to do with executable models may work, but are skeptical.
  • We believe that the goal of executable designs is exactly right and we think (along with many industry experts not employed by MS) that the DSL (domain specific language) approach is more likely to get us there.  This is a technical disagreement, not marketing hype.
  • Finally, we believe some customers will use both approaches.  UML, for instance, is much better at modeling arbitrary logical models than a DSL by nature.  UML today still has a better way for doing sequence designs (although we expect to address this in a DSL for services in v2).


Hopefully that provides a bit more clarity around  our DSL strategy as it relates to UML2.


Finally, I did have a chance to meet with some really interesting customers in 1-1 meetings in the afternoon.  What I heard basically validated for me the last 5 years of work.  Both said that the discussions around VSTS are the same discussions they’ve been having for the past year in terms of finding ways to improve their predictability of projects and decrease the costs.  Quality early and often resonates.  Services being designed not “created naturally in the dev process” was a major focus.  Understanding when and how to unit test and performance test were things they’ve been debating internally.  Overall, I learned a lot and found a couple more customers interested in piloting the Team System at Beta 2…